The pipes module is built around the RecursiveParserWrapper, and I'm happy with that being the default. I suspect that there are a number of users who will want the legacy behavior of getting a single metadata object + text for a document, no matter how many attachments. We recently did this on
Let's add a "parseMode" or similar to FetchEmitTuple to allow for the legacy behavior.
This will be a breaking change in the Solr and OpenSearch emitters. The "concatenate" and "skip" options for those emitters can now be handled via the fetchemituple/pipesiterator.